I am familiar with the debate surrounding Bates's decision to exclude p-values for mixed effects regression coefficients in lmer. However, I operate in a very p-value-focused discipline and am trying to clear something up.
What are the degrees of freedom for group-level predictors in a two-level model? I can use pvals.fnc() to obtain p-values for these coefficients but something tells me it is way overestimating the degrees of freedom for the tests, using the individual-level sample size and not the group-level sample size. Any help would be greatly appreciated.
